Int J Clin Exp Med 2017;10(10):14904-14910 www.ijcem.com /ISSN:1940-5901/IJCEM0059973

Original Article A convenient method for quantifying collagen fibers in atherosclerotic lesions by ImageJ software

Ying Chen1,2*, Qi Yu1,3*,Cang-Bao Xu1

1Shaanxi Key Laboratory of Ischemic Cardiovascular Disease & Institute of Basic and Translational Medicine, Xi’an Medical University, Xi’an, China; 2School of Electronic and Information Engineering, Xi’an Jiaotong University, Xi’an, China; 3Institute of Material Medical, School of Pharmacy, The Fourth Military Medical University, Xi’an, China. *Equal contributors. Received June 21, 2017; Accepted September 9, 2017; Epub October 15, 2017; Published October 30, 2017

Abstract: Background and aim: Collagen fibers are the main component of atherosclerotic lesions and should there- fore be used as an index to evaluate plaque stability. Although Masson’s trichrome stainis usually employed to identify the extracellular matrix, collagen fibers identified using this staining method are difficult to analyze because the three main colors often co-localize in same area within the atherosclerotic lesion. To solve this problem, we presented a convenient method for quantification of the collagen fibers in atherosclerotic lesions with Masson’s trichrome stain. Methods: This method is based on the open-resource image software ImageJ and an associated color deconvolution plugin. Original images of atherosclerotic lesions were converted into RGB images, and these images were deconvolved by ImageJ using the color deconvolution plugin. Results: The resulting monochromatic image obtained showed collagen fibers in maximal separation from background tissues. To calculate the area of the green component, collagen fibers were quantified in an accurate and efficient manner.Conclusions: Because of the simplicity and accuracy of this method, it may be widely applied for studying atherosclerosis in humans or various animal models.

Keywords: Collagen fiber, atherosclerotic lesion, smooth muscle cells, animal models

Introduction an atherosclerotic lesion, immunohistochemi- cal quantification of collagen fibers in an ath- The primary components of an atherosclerotic erosclerotic lesion is not practical because at lesion are extracellular matrix proteins, includ- least four distinct types of collagen have been ing collagen and elastin, which account for found. For this reason, numerous studies have more than 60% of the lesion [1]. Collagen used Masson’s trichrome stain to identify the degeneration results in atherosclerotic plaque collagen fibers regardless of collagen type [5, vulnerability; thus, the collagen content is con- 6]. However, with this three-color staining, col- sidered an indicator of plaque stability [2]. ors usually co-localize on the surface of tissue Notably, Shiomi et al. introduced the morpho- and strongly interfere with the identification of metric ‘vulnerability index’, which is the ratio of macrophages and extracellular lipid deposits to a specific component within the tissue [7]. If smooth muscle cells (SMCs) and collagen fibers the quantification of the collagen fibers relies in the atherosclerotic plaque [3]. This index is on the calculation of specific staining based on usually employed to evaluate the degree of visual inspection, this procedure could cause a plaque destabilization in animal models. Con- considerable number of errors. Therefore, to sidering that collagen fibers are responsible for overcome this obstacle, we developed a conve- the mechanical strength of the plaque, the nient method to quantify collagen fibers using number of collagen fibers determines the Masson’s trichrome stain and the open- mechanical properties of the plaque [4]. Unlike resource image software ImageJ with its color the identification of macrophages and SMCs in deconvolution plugin. Quantification of collagen in atherosclerotic lesions by ImageJ

in distilled water, they were treated with phos- phomolybdic acid for 10 min. Then, sections were stained with light green instead of Methyl Blue for 10 min to avoid any possible confusion between blue nuclear staining and collagen staining. In Masson’s trichrome stain, stained components were identified as the following: nuclei were stained black or blue; cytoplasm, muscle, and erythrocytes were stained red; and collagen fibers were stained green.

Quantification of collagen fibers by ImageJ

As shown in Figure 1, four steps are considered necessary for quantifying collagen fibers: instal- lation of the software and plugin, setting the scale, deconvolution of the color images and -quantifcation of the collagen fibers. The algo rithm described below was used in conjunction with version 1.8 of ImageJ.

Image processing Figure 1. Flow chart of the experimental Image acquisition is a prerequisite for further procedure. analysis, and four steps are required as previ- ously reported [11]. First, all images should be acquired under identical conditions. To avoid a Materials and methods short range on the grayscale (0-255), the auto- matic exposure and white balance should be Sample preparation turned off. Then, the blank part of images should be adjusted to white, and the false-pos- Japanese white rabbits were fed with a high- itive staining should be excluded from images. cholesterol diet (HCD; containing 0.3% choles- Last, all images should be digitized and stored terol and 3% corn oil) for 16 weeks to develop in an uncompressed tagged image file format atherosclerosis as previously described [8]. (tiff) with 24-bit RGB and 640 × 480 pixel After the rabbits were euthanized, the aortas resolution. were isolated and then fixed with 4% formalde- hyde solution. For histological analysis, the aor- Scale setting tic arch of each rabbit was cut into 10 cross sections (3 μm) as previously described [9]. As shown in Figure 2A, the scale bar was mea- Frozen cross sections of the aortic root in apoE sured by the “Straight” line tool in the “Analyze” deficient mice were kindly gifted by Dr. Hua menu, which allowed us to set the length of Guan (Research Institute of Atherosclerotic Di- scale bar in the “Known Distance” box and set sease, Xi’an Jiaotong University Cardiovascular the unit (μm) in the “Unit of Length” box under Research Center). The sections were stained the “Set scale” dialogue box. with Masson’s trichrome stain. Deconvolution of color images and identifica- Histological staining tion of collagen fibers

Masson’s trichrome stain was performed acc- Separation via color deconvolution provides a ording to a standard protocol [10]. Briefly, sec- means of separating collagen fibers from the tions were stained with Hansen’s iron hema- overlapping regions. The basis of this method is toxylin for 5 min. After the sections were wash- to separate the component stains by perform- ed under running tap water for 5 min, they were ing an orthonormal transformation of the stained with Biebrich scarlet-acid fuchsin for image’s RGB information [12]. During color 10 min. After the sections were further rinsed image deconvolution, the images were opened,

14905 Int J Clin Exp Med 2017;10(10):14904-14910 Quantification of collagen in atherosclerotic lesions by ImageJ

Figure 2. Procedure for processing images with ImageJ. Scale setting (A): 1) The scale bar was measured using the “Straight” line tool and 2) entered into the “Analy- sis” menu. 3) The length of scale bar was entered into the “Known Distance” box, and the unit (μm) was set in the “Unit of Length” box. (B) Color deconvolution: 1) The image was converted to an RGB color space and 2) processed using the “Color Deconvolution” plugin.

14906 Int J Clin Exp Med 2017;10(10):14904-14910 Quantification of collagen in atherosclerotic lesions by ImageJ

Figure 3. Images of atherosclerotic lesions processed by color deconvolution. The original RGB imagesof atheroscle- rotic lesions from rabbit (Light Green, upper), mouse (Light Green, middle) and rabbit (Methyl Blue, lower) were split into their red, blue and green components.

Figure 4. Quantification of the green component using ImageJ software. The threshold (A) was set to a constant value (B) for the green component. The results were shown as the area and integrated intensity (C). and the following steps were applied. First, the entering the “Image” menu, clicking on the input images were converted to RGB images by “Type” box, and choosing the “RGB Color” com-

14907 Int J Clin Exp Med 2017;10(10):14904-14910 Quantification of collagen in atherosclerotic lesions by ImageJ

Figure 5. Comparison of the image processing by RGB image splitting and color deconvolution. The split channels function separated the RGB image into red, blue and green channels (upper). The deconvolution plugin separated the RGB image into a red component, blue component and green component (lower). mand to convert the images (Figure 2B). atherosclerotic plaque was identified along the Second, images with Masson’s trichrome stain internal elastic lamina and the surface of the were processed using the “Color Deconvolu- plaque. tion” plugin by clicking “Color Deconvolution” in the “Plugins” menu to deconvolve the images Results into their red, blue and green components; the green component was identified as the colla- Collagen fibers were maximally separated from gen fibers Figures( 2B, 3). the background tissue

Quantification of collagen fibers As shown in Figure 3, the images of atheroscle- rotic lesions from both rabbits and mice were The area of the green collagen fibers was mea- separated into their red, blue and green com- sured after we entered the “Image” menu, click- ponents. However, if Methyl Blue was employed ed on the “Adjust” box, and isolated the green to stain collagen, the image was split into only area of collagen fibers using the “Threshold” its red and blue components (Figure 3). In the tool. The threshold was manually adjusted until blue component, blue staining of the nucleus the entire green area was highlighted in red made it difficult to quantify the collagen fibers. (Figure 4). Then, measurement of the threshold However, when light green was used to identify area was performed as follows: we entered the collagen fibers, a green image was obtained, set measurement dialog under the “Analyze” and the collagen fibers were shown with maxi- menu, and after we checked the “Area”, mal separation from the background tissue in “Integrated Intensity” and “Limit to Threshold”, the green component (Figure 3). when we clicked the “Measurement” button Collagen fibers were conveniently quantified by under the “Analyze” menu, the results were ImageJ presented in the “Results” window (Figure 4). Finally, area-based analysis was used to extract In the green image, the collagen fibers could be and quantify the regions of interest (ROIs) from quantified from the interested area Figure( 4). the image. This analysis allowed the perfor- When a constant threshold was set, the colla- mance of the algorithm to be evaluated. To gen fibers in different sections could be accu- quantify the collagen fibers in the ROI (i.e., the rately and consistently quantified by ImageJ atherosclerotic plaque), the “ROI manager” (Figure 4). To compare with staining separation was chosen in the “Tools” box under the into three channels, image with deconvolution- “Analyze” menu, and the region containing the indeed made the collagen fibers to obtain maxi-

14908 Int J Clin Exp Med 2017;10(10):14904-14910 Quantification of collagen in atherosclerotic lesions by ImageJ mal separation from the background tissue As above description, a green image was (Figure 5). obtained after image deconvolution, showing that collagen fibers were shown with maximal Discussion separation from the background tissue. In this study, we present a method to estimate Moreover, monochrome imaging also helps the area and integrated density of collagen researchers use this image for calculating the fibers in an atherosclerotic lesion; this method area and integrated intensity. Therefore, these is based on image analysis of a sample stained techniques offer the feasibility of automatically with Masson’s trichrome. As previously report- extracting collagen fiber data from atheroscle- ed, a wide range of collagen types are present rotic lesion images, which radically improve the in an atherosclerotic lesion, which increases methodology of collagen analysis based on the difficulty in quantifying collagens via immu- Masson’s trichrome stain. This method is more nohistochemical staining [12, 13]. Thus, stain- objective, sensitive and accurate compared to ing against various collagens within a lesion the traditional Masson’s trichrome staining may help researchers to better quantify the col- method. lagen density, and Masson’s trichrome stain is a potential option. In conclusion, we presented a convenient me- For more than eighty years, Masson’s trichrome thod to assess collagen fibers in atherosclerot- stain has been frequently used to stain colla- ic lesions using Masson’s trichrome stain. This gens in various tissues with either blue or method was based on the cross-platform soft- green, depending on the dyes available (aniline ware ImageJ and its associated color deconvo- blue or light green) [14]. However, because the lution plugin. Because of the simplicity and three main colors often co-localize to the same accuracy of this method, it may be widely area, it is difficult to analyze or quantify tissues applied for studying atherosclerosis in human stained with Masson’s trichrome stain [15]. In or various animal models. conventional digital imaging, bright field imag- Acknowledgements es are a composite of three 8-bit monochro- matic channels (red, green, and blue) that form This study was supported by the National a 24-bit color image [16]. Based on this conven- Natural Science of China (No. 81400328 and tion, the RGB images can be split into three 81773795) and China Postdoctoral Science channels (red, green, and blue), and using one Foundation (No. 2015M582800 and 2016T90- of these three channels can result in selective 972) and Natural Science Basic Research Plan contrast between common stains. Using this in Shaanxi Province of China (2016JM8025). separation technique, on a monochromatic image allows maximal separation between pos- Disclosure of conflict of interest itive-stained pixels and background tissue, which facilitates accurate calculations of both None. the area and intensity of immunohistochemical staining [17]. However, staining separation is Address correspondence to: Dr. Qi Yu, Institute of not suitable for analysis of Masson’s trichrome Basic and Translational Medicine, Xi’an Medical Uni- stain because this technique has inefficient versity, 1 Xinwang Road, Xi’an 710021, China. Tel: separation of the collagen fibers from the blue 86-29-82657057; Fax: 86-29-82657057; E-mail: or green staining of other tissues. As previously [email protected] reported, this incomplete separation was ascribed to the color spectrum of the available References stains, with the common colors partially spread over all three channels [7]. To overcome this [1] Adiguzel E, Ahmad PJ, Franco C and Bendeck problem, Ruifrok et al. developed a plugin MP. Collagens in the progression and compli- cations of atherosclerosis. Vasc Med 2009; named “color deconvolution”, which separates 14: 73-89. a three-channel image into three colors [18]. [2] Rekhter MD. How to evaluate plaque vulnera- Although some limitations should be consid- bility in animal models of atherosclerosis? ered in further applications, this technique can Cardiovasc Res 2002; 54: 36-41. effectively separate the contributions of two [3] Shiomi M, Ito T, Hirouchi Y and Enomoto M. co-localized stains [19]. Fibromuscular cap composition is important

14909 Int J Clin Exp Med 2017;10(10):14904-14910 Quantification of collagen in atherosclerotic lesions by ImageJ

for the stability of established atherosclerotic [12] Katsuda S, Okada Y, Minamoto T, Oda Y, Matsui plaques in mature WHHL rabbits treated with Y and Nakanishi I. Collagens in human athero- statins. Atherosclerosis 2001; 157: 75-84. sclerosis. Immunohistochemical analysis us- [4] Burleigh MC, Briggs AD, Lendon CL, Davies MJ, ing collagen type-specific antibodies. Arterios- Born GV and Richardson PD. Collagen types I cler Thromb 1992; 12: 494-502. and III, collagen content, GAGs and mechani- [13] Mostaco-Guidolin LB, Ko AC, Wang F, Xiang B, cal strength of human atherosclerotic plaque Hewko M, Tian G, Major A, Shiomi M and Sowa caps: span-wise variations. Atherosclerosis MG. Collagen morphology and texture analysis: 1992; 96: 71-81. from statistics to classification. Sci Rep 2013; [5] Clarke MC, Figg N, Maguire JJ, Davenport AP, 3: 2190. Goddard M, Littlewood TD and Bennett MR. [14] O’Connor WN and Valle S. A combination Apoptosis of vascular smooth muscle cells in- Verhoeff’s elastic and Masson’s trichrome duces features of plaque vulnerability in ath- stain for routine histology. Stain Technol 1982; erosclerosis. Nat Med 2006; 12: 1075-1080. 57: 207-210. [6] Shinohara M, Yamashita T, Tawa H, Takeda M, [15] Miot HA and Brianezi G. Morphometric analy- Sasaki N, Takaya T, Toh R, Takeuchi A, Ohigashi sis of dermal collagen by color clusters seg- T, Shinohara K, Kawashima S, Yokoyama M, mentation. An Bras Dermatol 2010; 85: 361- Hirata K and Momose A. Atherosclerotic plaque 364. imaging using phase-contrast X-ray computed [16] Brey EM, Lalani Z, Johnston C, Wong M, McIn- tomography. Am J Physiol Heart Circ Physiol tire LV, Duke PJ and Patrick CW Jr. Automated 2008; 294: H1094-1100. selection of DAB-labeled tissue for immun- [7] Tadrous PJ. Digital stain separation for histo- ohistochemical quantification. J Histochem logical images. J Microsc 2010; 240: 164-172. Cytochem 2003; 51: 575-584. [8] Yu Q, Li Y, Waqar AB, Wang Y, Huang B, Chen Y, [17] Vrekoussis T, Chaniotis V, Navrozoglou I, Zhao S, Yang P, Fan J and Liu E. Temporal and Dousias V, Pavlakis K, Stathopoulos EN and quantitative analysis of atherosclerotic lesions Zoras O. Image analysis of breast cancer im- in diet-induced hypercholesterolemic rabbits. J munohistochemistry-stained sections using Biomed Biotechnol 2012; 2012: 506159. ImageJ: an RGB-based model. Anticancer Res [9] Yu Q, Liu Z, Waqar AB, Ning B, Yang X, Shiomi 2009; 29: 4995-4998. M, Graham MJ, Crooke RM, Liu E, Dong S and [18] Ruifrok AC and Johnston DA. Quantification of Fan J. Effects of antisense oligonucleotides histochemical staining by color deconvolution. against C-reactive protein on the development Anal Quant Cytol Histol 2001; 23: 291-299. of atherosclerosis in WHHL rabbits. Mediators [19] Ruifrok AC, Katz RL and Johnston DA. Compari- Inflamm 2014; 2014: 979132. son of quantification of histochemical staining [10] Goldner J. A modification of the masson tri- by hue-saturation-intensity (HSI) transforma- chrome technique for routine laboratory pur- tion and color-deconvolution. Appl Immunohi- poses. Am J Pathol 1938; 14: 237-243. stochem Mol Morphol 2003; 11: 85-91. [11] Zhang G, Chen Y, BilalWaqar A, Han L, Jia M, Xu C and Yu Q. Quantitative analysis of rabbit cor- onary atherosclerosis. Practical techniques utilizing open-source software. Anal Quant Cytopathol Histpathol 2015; 37: 115-122.

14910 Int J Clin Exp Med 2017;10(10):14904-14910